42results about How to "Reliable input" patented technology

The invention belongs to the crossing field of artificial intelligence and barrier-free assistance technology, and particularly discloses a sign language translation method and device based on a large language model.The method is applied to a terminal and comprises the steps that key points are extracted from sign language video frames; based on the coordinates, types and global indexes of the key points in the sign language video frames, obtaining fusion features through feature fusion; based on the fusion features corresponding to the sign language video frames in the frame window, time sequence modeling and classification are carried out, and a word prediction result of the frame window is obtained; a sliding window voting and deduplication mode is adopted, the candidate word sequence is preprocessed, a preprocessed word sequence is obtained, and the candidate word sequence is formed by connecting word prediction results of all frame windows in series according to a time sequence; based on the preprocessed word sequence, word error correction and statement generation are carried out by calling a large language model, and a sign language translation result is generated. According to the invention, sign language video translation can be realized for the terminal, and the accuracy and fluency of the translation result can be guaranteed.

The application relates to an information processing method based on a large language model, a receiving method and related devices thereof, which comprises the following steps: obtaining an initial message to be processed, a target key and a current context; inputting the initial message to be processed into a target large language model, and obtaining a hash value corresponding to a current token in the initial message in a process of generating a token of a target text, and obtaining at least two candidate token sets according to the hash value; determining a target token set from the candidate token set according to the current token, and performing sampling control on candidate tokens in the target token set based on an original probability distribution of the current token to obtain a target token corresponding to a current generation position of the target text of the current token, and obtaining the target text. The scheme provided by the application can embed information by using a hash algorithm and a sampling control mechanism in the process of generating the text, and can improve the naturalness of the output text and the unbiasedness of the statistical distribution.

The application provides an engine active clearance design method, and belongs to the technical field of aero-engines. The method comprises the following steps: constructing a rigid-flexible coupling finite element model coupled with multi-body dynamics and finite elements; solving deformation under gyroscopic effect and inertial load and deformation under thermal coupling effect when the engine is in use; extracting clearance deformation between a rotating rotor and a rotating stator and clearance deformation between a rotating rotor and a stationary stator based on the solving of the deformation; identifying a deformation sensitive area; performing multi-objective optimization on the structure of the deformation sensitive area; based on the optimized result, and considering other factors that have an influence on the clearance between the rotor and the stator, designing an initial installation clearance value that takes into account safety and efficiency, wherein the other factors include radial movement of bearing and fulcrum elastic elements, rotor and stator runout, and relative axial position change of the rotor and the stator. Through the processing scheme, the accuracy of the active clearance design is improved, the safety and efficiency of the engine are taken into account, and the risk of engine collision and abrasion is reduced.

The application provides a kind of to identify batch electronic product average life evaluation method, it is related to the technical field of reliability engineering and life evaluation, solve the technical problem that average life evaluation method in prior art is difficult to accurately evaluate the average life of the batch electronic product.The method comprises establishing a segmented life distribution model, the segmented life distribution model includes a first stage life distribution model and a second stage life distribution model separated by a segmentation point time T, and the segmented life distribution model satisfies the parameter constraint relationship at the segmentation point time T;Wherein, the first stage life distribution model adopts an exponential distribution model with failure rate λ;The second stage life distribution model adopts a Weibull distribution model;Based on the segmented life distribution model, the average life of the batch electronic product is calculated and obtained by applying the average life analytical expression.The application can improve the calculation efficiency and calculation accuracy.

The invention relates to the technical field of electrochemical energy storage safety, and discloses a battery energy state estimation method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining an open-circuit voltage-state-of-charge curve of a battery; calculating an integral value of the open-circuit voltage-state-of-charge curve in a full-state-of-charge range; and estimating the real-time charge state of the battery at any moment, and calculating the energy state of the battery based on the integral value and the real-time charge state. According to the method, the open-circuit voltage-state-of-charge curve of the battery is obtained, the integral value of the curve in the full-state-of-charge range is calculated, and the energy state is calculated based on the integral value after the instant state-of-charge is estimated, so that the influence of internal resistance polarization on energy state estimation is effectively eliminated, and the precision and reliability of SOE estimation are remarkably improved; the problem of large error caused by internal resistance polarization is solved.

The application discloses a kind of fracture network multi-step iteration deep learning prediction methods for concealed rock mass, it is related to geological fracture structure intelligent identification and prediction technical field, including, extraction exposed rock area fracture network image, analysis exposed rock area in exposed rock mass fracture network image geometric characteristic parameter;Based on geometric characteristic parameter, exposed rock mass fracture network image is divided into different continuous sequence images, according to continuous sequence image construction training set and verification set;Deep learning neural network model is constructed, and training set and verification set are used to execute training process;Different image block size of continuous sequence image and adjacent sequence overlap rate are set as multiple candidate combinations, generate corresponding training set and verification set for each candidate combination, train and verify deep learning neural network model.Deep learning neural network model is constructed to predict the fracture network in concealed area, realize the continuous extrapolation of the fracture structure in masked area, and reduce the cumulative effect of prediction error.

This invention discloses a multimodal, multi-task spatial perception method for sanitation operations, comprising: acquiring LiDAR data and multi-view camera image sequences, constructing bird's-eye view representations for each, obtaining point cloud bird's-eye view features and image bird's-eye view features respectively; calculating fusion weights for the point cloud bird's-eye view features and image bird's-eye view features at the bird's-eye view grid positions based on the difference in semantic distribution between the two modalities and the confidence of each modality, and obtaining a shared bird's-eye view feature base through dynamic gating weighted fusion; performing target detection, semantic segmentation, and vector map detection on the shared bird's-eye view feature base respectively, obtaining the 3D bounding box parameters, semantic category probabilities, map element categories, and their point coordinate sequences of the target objects respectively; and correlating and cross-validating the 3D bounding box parameters, semantic category probabilities, map element categories, and their point coordinate sequences of the target objects.

The invention relates to the technical field of urban planning and meteorological disaster prevention, and particularly discloses an urban bad ventilation area diagnosis method based on a typhoon wind field-building group coupling effect. The method comprises the following steps: firstly, accurately simulating a typhoon wind field based on a WRF mode and an optimal urban canopy model; secondly, the optimal UCM scheme is embedded into the WRF, and city scale simulation under three-dimensional building parameter driving is achieved. Thirdly, constructing a high-quality boundary condition in combination with a classic wind profile function, driving block scale CFD simulation, and quantitatively evaluating the dynamic influence of the building density, height and layout on the near-earth wind field; and finally, proposing an urban form hierarchical optimization strategy based on a simulation result. According to the method, the bad ventilation area in the city can be diagnosed, the safety of the local wind environment is remarkably improved through an optimization strategy, the proportion of a static wind area is increased by 2-13%, the proportion of a strong wind area is reduced by 4-7%, and an effective tool is provided for typhoon resistance design of coastal cities.

This invention discloses a brain-computer interface (BCI) rehabilitation robot system for patients with neurological injuries, belonging to the field of BCI rehabilitation robot technology. The invention includes a multimodal signal acquisition module, an adaptive decoding module, and a collaborative execution module. By fusing flexible dry electrode EEG signals and near-infrared brain functional imaging signals, a Kalman filter model is used to perform weighted recursive estimation of time and spatial domain features, extracting high-confidence brain intention features. Based on a pre-trained general model, the model weights are dynamically fine-tuned to adapt to individual differences, and robust target behavioral feature parameters are generated based on confidence margin and historical state data. According to the behavioral feature parameters, active mechanical assistance state parameters and passive physiological stimulation state parameters are calculated in real time, driving the exoskeleton robot to perform precise motion assistance and functional electrical stimulation. This invention improves the personalization level and clinical efficacy of neurological injury rehabilitation training.

The present application relates to the technical field of geological data analysis, and discloses a bank slope deep unloading development zone identification and evaluation method and system based on a sliding window, aiming to solve the problems of low efficiency, strong subjectivity, discrete and non-quantifiable evaluation results of the existing method. The scheme mainly includes: obtaining the structured data of the fissures exposed along the exploration line; based on the set window parameters, performing spatial convolution calculation through the sliding window to generate a continuous characteristic signal sequence; using the preset background noise threshold to segment the sequence to identify the candidate section; matching and calibrating the boundary of the candidate section with the position of the known geological structure surface to delineate the deep unloading development zone; and calculating the comprehensive development intensity index of each deep unloading development zone to realize quantitative evaluation. The present application realizes the automatic, continuous spatial identification and quantitative characterization of the deep unloading development zone, improves the efficiency and accuracy, and is particularly suitable for the safe site selection and construction of major projects.

The application relates to a large-scale underwater explosion problem, in particular to a calculation method of underwater explosion load based on an LDGM and MPI. The calculation method of underwater explosion load based on the LDGM and MPI comprises the following steps: performing non-structural grid division on a water area geometric model, calculating data in the non-structural grid after spatial discretization and time discretization according to a LDGM second-order wave equation, performing regional decomposition by using METIS during the calculation process, dividing the non-structural network into multiple subdomains and distributing the subdomains to different processors, exchanging numerical flux on the subdomain boundaries between the processors, performing parallel processing, until time domain loop calculation is completed, and obtaining a pressure field of a flow field evolving with time. The LDGM is used to solve the flow field pressure load, can more accurately capture the strong gradient of a shock wave and local details of a cavitation area, avoids numerical dissipation and oscillation problems, and improves calculation efficiency through a parallel processing scheme.

The present application relates to the technical field of data prediction, and discloses a coal calorific value high-precision prediction method based on machine learning; the method comprises the following steps: obtaining coal quality detection data and performing windowing processing to form a coal calorific value sequence and a multi-source coal quality component sequence; a disturbance embedding vector group is constructed according to adjacent measurement value changes; statistical features and dominant amplitudes of coal quality component variables are extracted, a signal-to-noise factor is constructed, and a selection weight is obtained through mapping to generate a variable-level coal quality feature vector group; the disturbance embedding vector group and the variable-level coal quality feature vector group are combined in time sequence to form a coal quality correlation modeling set, and a sparse correlation structure field is constructed based on a training set; a comprehensive state representation is obtained through correlation propagation and steady-state consistency constraints; a test set is input into the mechanism to obtain a coal calorific value prediction result sequence; the method can strengthen disturbance expression, improve feature fusion quality, and maintain prediction structure stability, and a coal calorific value high-precision prediction result is obtained.

The application belongs to the technical field of intelligent breeding, and provides a sow leg automatic scoring method and device based on three-dimensional point cloud. The method comprises the following steps: acquiring a target three-dimensional point cloud model of a pig; extracting a leg point cloud set corresponding to each leg of the pig from the target three-dimensional point cloud model; determining leg key point data of each leg from the leg point cloud set corresponding to each leg; determining trunk key point data from the target three-dimensional point cloud model according to the leg key point data; determining a posture determination result of the pig according to the target three-dimensional point cloud model, the leg key point data of each leg and the trunk key point data; and in the case that the posture determination result indicates that the pig is in a standard standing posture, determining a leg structure type and a leg structure score of the pig according to the leg key point data. The application can stably identify the leg structure characteristics of the pig and realize consistent automatic scoring, and has good application potential in terms of accuracy, robustness and interpretability.

The invention provides a muck pile ground point cloud removing method and device, and relates to the technical field of underground metal mine blasting mining engineering. The method comprises the steps of obtaining original three-dimensional point cloud data of a muck pile and performing down-sampling to obtain a muck pile three-dimensional point cloud; determining a segmentation threshold according to the feature distribution of the muck pile three-dimensional point clouds, and performing segmentation by adopting a region growing algorithm to obtain a plurality of segmentation result point clouds; calculating and screening geometric feature vectors of each segmentation result point cloud; based on the screened segmentation result point cloud feature data set, ground class feature subsets are extracted through unsupervised clustering analysis, a ground point cloud joint judgment rule is constructed, and each segmentation result point cloud is judged as a ground point cloud or a non-ground point cloud; and all identified ground point clouds are removed, and all non-ground point clouds are subjected to denoising processing to obtain muck pile main body point clouds. The invention aims to improve the quality and availability of the three-dimensional point cloud data of the muck pile and ensure efficient and accurate identification and statistics of the muck pile lumpiness.

This application provides an exposure parameter adjustment method, apparatus, electronic device, and storage medium. The method includes: acquiring a monitoring image of the furnace mouth area of ​​a smelting furnace, and extracting a mask image of the furnace mouth flame from the monitoring image. The monitoring image is acquired by an industrial camera. Based on the monitoring image and the mask image, the method calculates a first-image brightness of the flame area and a second-image brightness of the effective area. Based on the first-image brightness and the second-image brightness, the method calculates the current image brightness of the flame area. The effective area is the region in the monitoring image whose grayscale value is greater than or equal to a preset grayscale threshold. The method calculates the brightness deviation between the current image brightness and a preset target image brightness, and adjusts the exposure parameters of the industrial camera based on the brightness deviation. By combining the brightness characteristics of the flame area in the monitoring image and the brightness characteristics of the overall effective area, the exposure parameters of the industrial camera are dynamically adjusted, ensuring the quality stability of the video image of the flame in the furnace mouth flame monitoring.

The invention belongs to the technical field of engines, and particularly relates to a diesel engine in-cylinder heat release curve real-time reconstruction method based on injection parameter variation, which comprises the following steps of: extracting diesel engine structure, control and operation parameters from a steady-state engine characteristic database, determining a typical load working condition and obtaining a reference combustion heat release rate curve; the circulating air inflow is calculated based on the rotating speed of the diesel engine and the pressure of an air inlet main pipe, the in-cylinder excess air coefficient is calculated according to the current circulating air inflow and the oil injection amount, the injection duration and the combustion duration are calculated in combination with the rotating speed of the diesel engine, the injection pressure, the circulating oil injection amount and the excess air coefficient, and the reference combustion heat release rate is further calculated. And phase translation is carried out on the reconstructed combustion heat release rate curve according to the main injection advance angle, and a target combustion heat release rate curve consistent with the main injection moment is obtained. The method is small in calculation amount and high in reconstruction speed, can adapt to the influence of injection parameter change on the combustion heat release characteristic under the transient working condition, and meets the application requirements of real-time analysis and vehicle-mounted control.

The invention belongs to the technical field of meteorological simulation, discloses a systematic tropical cyclone and high-resolution wind field modeling method, system and device based on environment driving and a medium, and aims to solve the problem of wind field simulation distortion in complex terrains. The method comprises the following steps: acquiring and unifying environment field data; calculating a dynamic generation potential index field based on the base month average climate field, defining a space-time generation probability, and randomly sampling to generate an initial event set of a plurality of synthetic tropical cyclone events; performing iterative simulation on each synthetic tropical cyclone event, and outputting full life cycle information; aiming at the full life cycle information, based on a gradient wind balance physical model, calculating each position point in the moving track information and a corresponding central air pressure difference value, and obtaining a near-surface wind field under an open terrain assumption as a first near-surface wind field; and acquiring high-resolution terrain data, inputting the high-resolution terrain data and the first near-surface wind field into the terrain adaptive correction model, and outputting the corrected near-surface wind field as a second near-surface wind field.

A power distribution network fault section positioning method considering pre-disaster reinforcement information comprises the following steps: step 1, establishing a first-stage pre-disaster reinforcement model used for determining a reinforcement scheme of a feeder line section and FTU nodes, calculating corresponding investment cost, and generating a fault scene set based on the reinforcement scheme; step 2, establishing a second-stage section positioning model for fusing multi-source information to perform fault section identification and calculating positioning deviation penalty cost; 3, integrating the pre-disaster reinforcement model in the step 1 and the second-stage section positioning model in the step 2, and constructing a two-stage collaborative optimization model with the aim of minimizing the total expected cost of the system; and 4, solving the two-stage collaborative optimization model constructed in the step 3, and finally outputting an optimal pre-disaster reinforcement scheme. According to the method, through collaborative optimization of pre-disaster reinforcement and fault positioning, the fault positioning fault-tolerant capability and the system toughness of the power distribution network under extreme disasters are remarkably improved.

The invention belongs to the field of new energy, relates to a charging pile energy efficiency optimization technology, is used for solving the problem that in the prior art, power control logic based on isolated voltage sensing has fundamental defects when coping with group interaction behaviors in a multi-pile coexistence scene, and particularly relates to a charging pile energy efficiency self-optimization system based on multi-modal AI. Comprising a data acquisition module, a feature calculation module, an intermediate state mapping module, a dynamic modulation module, a power adjustment amount calculation module and an execution and broadcast module. According to the method, the linear fitting in the preset time window is carried out on the real-time power sequence of each neighbor charging pile, the power change rate of each neighbor charging pile is accurately calculated, the number of the charging piles with the power change being the positive value and the negative value is further counted, and the consistency of the neighbor charging pile group in the power adjustment direction can be accurately quantified.

The invention provides a size precision regulation and control method and system for micro-jet 3D printing, and the method comprises the steps: obtaining technological parameters and corresponding formed part parameters, and constructing a collaborative optimization model according to a mapping relation between the technological parameters and the corresponding formed part parameters; a target structure model is obtained, the corresponding printing powder particle size is determined, the layering thickness is determined according to the target structure model and the printing powder particle size, and jet parameters and density parameters are determined through the collaborative optimization model based on the printing powder particle size and the layering thickness; according to the particle size of the printing powder, the layering thickness, the density parameter, the injection parameter and the target structure model, a local shrinkage rate is determined through a preset shrinkage rate prediction function so as to determine a corresponding local reverse compensation coefficient, and then according to the local reverse compensation coefficient, local size adjustment is conducted on the target structure model to determine a correction model; and printing according to the process parameters and the correction model. The problem that in the prior art, a high-universality, high-precision and efficient size precision regulation and control method for micro-jet 3D printing is lacked is solved.

This invention provides a quantum computing method based on fuzzy processing, relating to the field of quantum computing technology. The method includes: inputting the fuzzy quantum state into a quantum circuit containing at least one quantum logic gate operation to obtain an output quantum state; performing continuous weak measurements on the qubits during the operation of the quantum circuit to obtain imprecise error information; processing the imprecise error information based on fuzzy mathematics to obtain error estimation parameters; generating and applying feedback operations to the qubits based on the error estimation parameters to suppress errors; measuring the output quantum state and determining the output result according to a preset probability threshold interval, where the probability threshold interval is used to distinguish different output categories. This invention can also be further combined with quantum error correction to achieve long-depth quantum computing using error correction methods, realizing effective quantum state stability control.

The invention discloses a radar-oriented order constraint Jacobi eigenvalue self-sorting method and system, solves the problem of resource consumption caused by an additional sorting module in a Jacobi iteration eigenvalue decomposition post-processing flow in the prior art, realizes direct output of ordered eigenvalues and eigenvectors, and omits an independent sorting module. The method comprises the steps that a real symmetric matrix A obtained through radar baseband signal sampling and covariance estimation is processed to obtain a compressed array, and a feature vector matrix V is initialized; in the parallel iteration process, index pairs are generated according to a scheduling sequence, sub-matrixes are extracted, two types of asymmetric rotation operators are generated through a condition driving mechanism, and a feature vector matrix is synchronously updated to track feature vectors; global data replacement is carried out, and non-diagonal element energy judgment convergence is calculated; after iteration is ended, feature values arranged in a descending order can be directly extracted from the main diagonal of the final compressed array, and corresponding feature vectors are output and used for direction of arrival estimation.

The invention discloses an automatic positioning and direction finding method and device based on AISG interface transmission. The method comprises the following steps: S1, device composition and deployment; s2, parasitic power supply; s3, multi-source data collaborative acquisition; s4, data preprocessing and precision optimization; s5, cloud-edge cooperative positioning calculation is carried out; s6, AISG interface time division multiplexing transmission is carried out; s7, data automatic application and abnormity alarm are carried out; according to the invention, the device can be quickly attached and fixed to the base station antenna backboard without modifying the antenna through the flexible housing by using the viscous structure; the LDO voltage stabilization module multiplexes an RRU interface and is matched with a super capacitor to prevent power failure, data is subjected to temperature drift compensation and Kalman filtering to improve the direction finding precision, RTK calculation of CORS data is combined to guarantee accurate positioning, and multiple modules cooperate to stably output the data; the AISG time division multiplexing link does not affect the original function of the antenna, data is in seamless connection with a base station system, automatic input and abnormal alarm are achieved, and automatic operation and maintenance are adapted.

The invention belongs to the field of green traffic, and particularly relates to a microcosmic traffic simulation speed trajectory regeneration method responding to emission evaluation. Comprising the following steps: 1, selecting an emission factor model, and simulating and outputting second-by-second speed, lane and position data of a vehicle; step 2, constructing acceleration-jerk envelope line models of different vehicle types; 3, considering constraint conditions of kinematics consistency, dynamics controllability, jerk boundary, following safety and stroke boundary consistency, and constructing a vehicle speed trajectory control model; 4, performing linearization processing on nonlinear constraints in the vehicle speed trajectory control model by using an SOS2 technology; and 5, designing a'relaxation-strengthening 'solving algorithm, and optimizing output to obtain a speed trajectory sequence responding to emission evaluation and physical consistency. According to the method, a transferable technical path is provided for a robust coupling microscopic simulation and emission model, and the method can be used for hot spot analysis of a project scale, near-path air quality evaluation and design of a refined emission reduction strategy.

The application belongs to the technical field of construction engineering, and particularly relates to a building bearing pressure uplift pile differential settlement double-path grouting collaborative control system. The system constructs a construction time sequence data set covering geology, design and real-time monitoring through a collection module; a displacement prediction module is used to input data into a dynamic parameter corrected soil squeezing effect prediction model and a differential settlement prediction model, and output accurate soil squeezing displacement and differential adjacent pile body settlement difference prediction values; a parameter optimization module generates a construction control instruction containing an optimized construction parameter benchmark value based on the prediction values and a preset threshold value; a control feedback module generates a control instruction to drive on-site equipment to execute by calculating the deviation and dynamically adjusting the parameters through real-time collection of dynamic feedback data, and finally realizes collaborative dynamic control of three key targets of soil squeezing effect, differential settlement and joint waterproofing, effectively guaranteeing the construction quality and safety of complex pile foundation engineering.

The application relates to the field of Bluetooth positioning technology, in particular to a Bluetooth positioning method and system based on a logarithmic distance path loss model, which comprises the following steps: acquiring signal strength indication data of each Bluetooth access point at multiple reference points in a target area; fitting path loss parameters of the logarithmic distance path loss model according to the signal strength indication data at each reference point and the actual distance from the corresponding reference point to each Bluetooth access point; generating a reference signal strength matrix based on the signal strength values collected at each reference point, acquiring the signal strength values of each Bluetooth access point at a target point to be positioned, and forming a target signal strength vector. The application generates weights based on error reciprocals, performs weighted average on the measured signal strength of neighbor reference points, obtains compensated signal strength values, effectively suppresses the interference of indoor environment multipath effect, signal fluctuation and path loss model parameter fitting deviation on signal strength estimation, and makes the compensated signal strength closer to the real value.

This invention belongs to the field of radar array signal blind source separation technology, specifically involving an adaptive blind separation method, program, and system for radar main lobe composite interference signals. This invention constructs a robust complex whitening matrix using a non-zero time delay covariance matrix, reducing the complete dependence on the zero time delay covariance matrix during whitening. This reduces the impact of spatial colored noise on the whitening matrix and improves the stability of blind separation processing under spatial colored noise conditions. This invention uses the number of time delay covariance matrices participating in joint diagonalization as an adaptive selection object, enabling the determination of the target time delay matrix number based on changes in the second-order correlation structure under different composite interference scenarios, avoiding scenario mismatch problems caused by fixed number settings. This invention combines robust complex whitening, adaptive time delay matrix number selection, and joint diagonalization demixing processing into a unified second-order blind identification process, enabling the separation of radar main lobe composite interference observation signals into clearer signal components.

The invention discloses a method for calculating static horizontal force of a contact network locator, which comprises the following steps: a rectangular coordinate system is established by taking a locating point A of the locator as an original point O, the locator applies static horizontal force to a contact line, and the positive direction of a Y axis is consistent with the direction of the horizontal force; the contact line is in corner deformation relative to the positioning point to form a deflection line part, the deflection line AB and the deflection line AC are symmetrical relative to the positioning point, and the outward extending parts of the two points B and C on the contact line are straight line parts; carrying out stress analysis on the deflection curve part of the contact line, obtaining a control differential equation according to a deflection curve differential equation of the tension beam, and solving to obtain a control differential equation solution; the deflection curve parts of the contact lines on the two sides of the positioning point A are approximated to be axial force cantilever beams, the axial force cantilever beam bending moment is obtained according to axial force cantilever beam bending moment balance, and therefore the horizontal force generated at the positioning point A due to the bending rigidity of the contact lines is obtained; the method improves the calculation precision of the horizontal force of the positioning point.

This application discloses a hierarchical adaptive intelligent association and alignment method for multimodal flight data, comprising: acquiring multi-source airborne flight data and preprocessing it; mapping data of different modes in the preprocessed multi-source airborne flight data to a common feature space to obtain cross-modal alignment features, wherein an aircraft physical dynamic model is introduced as a constraint during the mapping process; performing local precise association and global semantic association in parallel based on the cross-modal alignment features; wherein the local precise association is used to obtain pixel-level or point-level matching results, and the global semantic association is used to obtain globally consistent semantic association results; and fusing the results of the local precise association and the global semantic association to obtain association information of corresponding granularity. This invention effectively solves the "data silo" problem caused by differences in format, spatiotemporal reference, and modality of multi-source flight data, significantly improving association accuracy, robustness, and system adaptability.

The invention relates to the field of wind power generation, in particular to a complex terrain wind power plant power and load collaborative optimization control method and system. Acquiring operation data and environment data of the target wind power plant in real time; constructing a multi-fan wake flow influence evaluation model of the target wind power plant based on the environmental data; establishing a multi-objective optimization function by taking the maximization of the total generated power of the wind power plant as a first optimization objective and the minimization of the load of the key structural component of each wind turbine generator as a second optimization objective; on the basis of a multi-fan wake flow influence evaluation model, predicting equivalent input wind speeds of all the wind turbine generators under different control strategies; according to the method, maximization of the total generated power of the wind power plant and minimization of the loads of key structural components of all wind turbine generators serve as optimization objectives, a multi-objective optimization function is established, it is guaranteed that the generating capacity is improved, meanwhile, the fatigue loads of a tower, blades and other components can be reduced, the service life of a draught fan is prolonged, and the operation and maintenance cost is reduced; and the operation safety and economy of the wind power plant are improved.